Matplotlib logarithmic scatter plot

[Derek Basch]

Can anyone give any suggestions on how to make a logarithmic (base 10)
x and y axis (loglog) plot in matplotlib? The scatter function doesn't
seem to have any log functionality built into it.

Thanks,
Derek Basch

P.S. I suck at math so feel free to make me feel stupid if it is really
easy to do .

 

[Bas]

Try this, don't know if this works for al versions:

from pylab import *
x=10**linspace(0,5,100)
y=1/(1+x/1000)
loglog(x,y)
show()

If you only need a logarithm along one axis you can use semilogx() or
semilogy(). For more detailed questions go to the matplotlib mailing
list.

Cheers,
Bas

 

[Derek Basch]

Thanks for the reply. I need a scatter plot though. Can that be done?

 

[John Hunter]

Derek> Thanks for the reply. I need a scatter plot though. Can
Derek> that be done?

You can set the scale of xaxis and yaxis to either log or linear for
scatter plots

In [33]: ax = subplot(111)

In [34]: ax.scatter( 1e6*rand(1000), rand(1000))
Out[34]: <matplotlib.collections.RegularPolyCollection instance at
0x9c32fac>

In [35]: ax.set_xscale('log')

In [36]: ax.set_xlim(1e-6,1e6)
Out[36]: (9.9999999999999995e-07, 1000000.0)

In [37]: draw()

 

[Derek Basch]

Great! That worked fine after I played with it for a bit. One last
question though. How do I label the ticks with the product of the
exponentiation? For instance:

100

instead of

10**2

Thanks for all the help,
Derek Basch

 

[John Hunter]

Derek> Great! That worked fine after I played with it for a
Derek> bit. One last question though. How do I label the ticks
Derek> with the product of the exponentiation? For instance:

Derek> 100

Derek> instead of

Derek> 10**2

You can supply your own custom tick formatters (and locators). See

http://matplotlib.sf.net/matplotlib.ticker.html

and examples

http://matplotlib.sourceforge.net/examples/custom_ticker1.py
http://matplotlib.sourceforge.net/examples/major_minor_demo1.py
http://matplotlib.sourceforge.net/examples/major_minor_demo2.py

JDH

 

[Derek Basch]

Thanks again. Here is the finished product. Maybe it will help someone
in the future:

from pylab import *

def log_10_product(x, pos):
"""The two args are the value and tick position.
Label ticks with the product of the exponentiation"""
return '%1i' % (x)

ax = subplot(111)
# Axis scale must be set prior to declaring the Formatter
# If it is not the Formatter will use the default log labels for ticks.
ax.set_xscale('log')
ax.set_yscale('log')

formatter = FuncFormatter(log_10_product)
ax.xaxis.set_major_formatter(formatter)
ax.yaxis.set_major_formatter(formatter)

ax.scatter( [3, 5, 70, 700, 900], [4, 8, 120, 160, 200], s=8, c='b',
marker='s', faceted=False)
ax.scatter( [1000, 2000, 3000, 4000, 5000], [2000, 4000, 6000, 8000,
1000], s=8, c='r', marker='s', faceted=False)

ax.set_xlim(1e-1, 1e5)
ax.set_ylim(1e-1, 1e5)
grid(True)
xlabel(r"Result", fontsize = 12)
ylabel(r"Prediction", fontsize = 12)

show()

 

[John Hunter]

Derek> formatter = FuncFormatter(log_10_product)
Derek> ax.xaxis.set_major_formatter(formatter)
Derek> ax.yaxis.set_major_formatter(formatter)

I would caution you against using identical objects for the x and y
axis *Locators*. For the formatters, it will do no harm, but for the
locators you can get screwed up because the locator object reads the
axis data and view limits when making it's choices.

Ie, do not do this:

ax.xaxis.set_major_locator(locator)
ax.yaxis.set_major_locator(locator)

but rather do this

ax.xaxis.set_major_locator(MyLocator())
ax.yaxis.set_major_locator(Mylocator())

Thanks for the example,
JDH

 

[Derek Basch]

Good tip John. Hopefully it will help someone out. Thanks again.
Derek Basch

 

[evander21]

You can also keep plot() from connecting the points

In the process of trying to solve what I think is the same problem I ran across a potential more convenient and elegant solution. The following command effectively creates a loglog scatter plot:

import matplotlib.pyplot as plt

plt.loglog(x,y,marker='o',linestyle='none')

hope this helps,

evan